SD3 ControlNet support + New ComfyUI Compatibility #123
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…to avoid conflicts with upcoming ComfyUI changes
…k on making latent preprocs obsolete, still need to make control weights apply the same way as before
…Ctrl and ReferenceCN
…sed feature regression at some point)
… T2IAdapter softweights have breaking changes due to extra middle weight)
…e would be with SDXL T2IAdapter models with custom/soft weights), automatically resize T2IAdapter control tensors to match batch_size to make my life easier
Kosinkadink
changed the title
|
ControlNetApplysD3 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Not backwards compatible - will not work with any ComfyUI version prior to the most recent that adds SD3 controlnet support. Changes are too drastic to reasonably keep backwards compatibility.
Adds SD3 support by adding a vae_optional input to Apply Advanced ControlNet node - soon, will be able to replace latent preprocs for reference cn/sparse rgb.
Only thing that might produce different results currently is SDXL T2IAdapter models when used with custom/soft weights - the middle weight is handled differently than before, would need to do additional work to make it work identical to before.